The GSE98793 and GSE16561 datasets were screened from the GEO database using the keywords (depression and peripheral blood, ischemic stroke and peripheral blood), respectively. The GSE98793 dataset contains whole blood samples from 64 MDD and 64 CON, and the GSE16561 dataset contains peripheral blood samples from 39 IS patients and 24 CON.

The “limma” package in R software was used to screen the DEGs between MDD and CON, IS and CON in the two datasets (p <0.05, |logFC|>0.1), respectively. It is then visualized with heat maps and volcanic maps. Finally, the up-regulated and down-regulated genes in the two datasets were intercrossed by Venn diagram to obtain the differential genes related to both diseases.

The DEGs were analyzed by KEGG pathway enrichment and GO analysis. GO analysis includes three categories: biological process (BP), cell component (CC) and molecular function (MF).

PPI network analysis of DEGs was performed using the STRING online website (https://string-db.org/). Then the results from STRING database were imported into Cytoscape for visualization, and cytoHubba was used to identify the hub genes in the PPI network. The top 10 genes were identified by four algorithms, including “Degree”, “EPC”, “MCC” and “MNC”, and then take the intersection to get the candidate genes.

The two datasets were analyzed using RF algorithm to screen the top 20 genes, and then the selected genes were interleaved with the genes obtained in PPI to obtain candidate hub genes.

The expression of candidate hub genes in both datasets was analyzed, and then ROC analysis was performed on both datasets to evaluate the sensitivity of the candidate hub gene.

The string database was used to find the genes that interacted with the candidate hub genes. RNAInter (http://www.rnainter.org/) database, miRWalk (http://mirwalk.umm.uni-heidelberg.de/) database and miRDB (https://mirdb.org/) database were used to screen the miRNA that interacted with the candidate hub genes, and then Venn diagram was used to find out the co-interacting miRNA. Starbase was used to identify the circRNA interacting with the predicted miRNA, and then Cytoscape was used to visualize the ceRNA network of the interacting circRNA-miRNA-mRNA.

We screened the candidate drugs that target the candidate hub genes using DSigDB (http://tanlab.ucdenver.edu/DSigDB) database.

Based on gene expression in each sample, the “CIBERSORT” package in R software was used to analyze the proportion of 22 immune cells in each sample. The differences in immune cells among different groups were calculated and visualized. The expression of immune cells and immune checkpoint was analyzed. We also analyzed the correlation between the expression of the candidate hub gene and immune checkpoint.

All participants (40-70 years old) underwent an initial clinical assessment, including the collection of clinical and demographic information. Depression symptoms in post-stroke patients were evaluated by the Hamilton Depression Rating Scale 17-item (HAMD-17) at one month after stroke by a trained neurologist. A score of 0–7 was considered normal, while a HAMD score > 7 is indicative of depression. Stroke severity was measured using the National Institute of Health Stroke Scale (NIHSS). All MDD patients were hospital-diagnosed according to the International Classification of Disease 10. The inclusion criteria for MDD are specified below: (1) age, 40–70 years; (2) HAMD-17 scores > 7; (3) MDD patients consisted of two scenarios, a first episode not previously treated with any antidepressant medication, or a history of MDD with symptomatic relief on antidepressant medication and cessation of any antidepressant medication for at least one month. The exclusion criteria for all participants are specified below: (1) pregnancy and pregnant females; (2) serious other neurological diseases and tumors, kidney and cardiovascular diseases; (3) participants with any signs of infection. Whole blood samples were collected upon admission. Plasma was then extracted and stored at -80°C.

This study was approved by the Research Ethics Committee of the Affiliated Nanjing Brian Hospital, Nanjing Medical University. Written informed consent was acquired from all participants.

The plasm RNA from patients and CON was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA, USA). The synthesis of cDNA was achieved using HiScript IV RT SuperMix for qPCR (R423-01, Vazyme). The Taq Pro Universal SYBR qPCR Master Mix (Q712-02, Vazyme) was used to detect the expression of cDNA.

R software (v4.3.1) was used for all statistical analysis. The differences among four groups were evaluated using one-way analysis of variance (ANOVA) followed by Tukey post hoc tests, T-tests and χ2-tests. The quantitative data are presented as means ± SEM using Graphpad Prism9 or SPSS20.0 software (SPSS, Inc., Chicago, IL, USA). The p-value < 0.05 was considered statistically significant.

The mRNA expression of the two data sets is shown in Figure 1A . In GSE98793, 1204 genes were upregulated and 673 genes were downregulated in MDD compared to the CON ( Figure 1B ). 1684 down-regulated genes and 1371 up-regulated genes were screened in GSE16561 ( Figure 1B ). Then the significantly up-regulated and down-regulated genes of the two datasets were intersected, respectively, and a total of 241 up-regulated genes and 153 down-regulated genes were screened ( Figure 1C ).

GO analysis of the DEGs was showed in Figure 2A . The BP category was enriched in activation of immune response, immune response-regulating signaling pathway, immune response-activating signaling pathway, positive regulation of cytokine production, regulation of response to biotic stimulus, regulation of innate immune response, blood coagulation, coagulation, hemostasis, and regulation of reactive oxygen species metabolic process. For the CC, enrichment was seen in specific granule, secretory granule lumen, cytoplasmic vesicle lumen, vesicle lumen, secretory granule membrane, tertiary granule, specific granule lumen, primary lysosome, specific granule membrane and tertiary granule lumen. The MF for these DEGs was glucose integrin binding. KEGG analyses revealed that the DEGs were involved in hematopoietic cell lineage, and complement and coagulation cascades ( Figure 2B ).

To explore core genes from the DEGs, the DEGs were uploaded to STING database for PPI network analysis ( Figure 3A ), and then the STRING results were imported into Cytoscape for visualization ( Figure 3B ).

Four algorithms (Degree, EPC, MCC and MNC) in Cytoscape were utilized to calculate the hub genes ( Figures 3C–F ). CD8A, IL7R and GZMA were identified as the potential hub genes ( Figure 3G ).

RF algorithm was used to screen the intersection of the top 20 potential genes of the two datasets, respectively ( Figures 4A, B ). And then, the intersection genes obtained by RF algorithm and the three candidate genes mentioned above were intersected, leading to the identification of IL-7R as a potential diagnostic marker ( Figures 4C, D ).

IL-7R expression levels were significantly lower in MDD patients than in CON in GSE98793 dataset ( Figure 5A ). And IL-7R expression levels also were significantly lower in IS patients than in CON in GSE16561 dataset ( Figure 5B ).

Based on these two data sets, ROC analysis was then performed on IL-7R. The results showed that IL-7R had a good diagnostic and predictive effect on MDD patients and IS patients, with AUC of 0.7332 of GSE98793 and 0.9081 of GSE16561, respectively ( Figures 5C, D ).

We constructed the IL7R-related gene-gene interaction network using String database, and the results revealed that 10 central genes are connected to IL-7R ( Figure 6A ).

Using the RNAInter, miRWalk and miRDB, we found 9 miRNA that can interact with IL-7R, including hsa-miR-204-5p, hsa-miR-211-5p, hsa-miR-32-3p, hsa-miR-3682-3p, hsa-miR-5197-3p, hsa-miR-6760-3p, hsa-miR-6808-3p, hsa-miR-6881-3p and hsa-miR-4764-3p ( Figure 6B ). Then we used ENCORI database to explore the circRNA that can regulate miRNA and the circRNA-miRNA-mRNA regulatory network was constructed by Cytoscape software ( Figure 6C ).

We found the perturbagen signatures and computational drug signatures that can target IL-7R, which shown in Supplementary Tables S1 , S2 (In the Supplementary Materials ) by DSigDB database. In total, we found 7 and 22 drugs that could increase and decrease IL-7R expression in Supplementary Table S1 . In addition, there are 32 drug signatures extracted from literatures using a mixture of manual curation and by automatic computational approaches in Supplementary Table S2 .

The CIBERSORT was used to analyze the proportion of 22 types of immune cells in peripheral blood from two datasets. In GSE98793, the proportions of macrophages M0 and monocytes in MDD patients were significantly higher than those in CON. However, the proportion of eosinophils, T cells CD4 memory resting and T cells gamma delta in MDD was significantly lower than that in CON ( Figures 7A, B ). Next, we analyzed the differences in immune checkpoint gene expression between MDD patients and CON. Compared with CON, there are significant differences 11 gene expression, including 4 high expressed genes (SIRPA, HLA - E and LGALS9, etc.) and 7 lower expressions (CD226, TIGIT and CD96, etc., Figure 7C ). Finally, the correlation between IL-7R and the above immune checkpoint genes was analyzed. The results showed that IL-7R was positively correlated with BTLA, CD226, CD28 and CD96, and negatively correlated with HLA-E, LGALS9, PVR and SIRPA ( Figure 7D ).

In GSE16561, we found that Dendritic cells activated, Macrophages M0, Monocytes, and Neutrophils were the most significantly increased, and NK cells resting, T cells CD4 naïve and T cells CD6 were the most significantly decreased in IS patients, compared with CON ( Figures 8A, B ). The differences in immune checkpoint gene expression between MDD patients and CON were shown in Figure 8C . The correlation between IL-7R and the above immune checkpoint genes were presented in Figure 8D . There are 19 immune checkpoint genes are positively correlated with IL-7R and 3 immune checkpoint genes are negatively with IL-7R.

The demographic and clinical characteristics of patients and CON are summarized in Supplementary Table S3 . There were no significant differences in age (F = 1.416, p _ANOVA = 0.248), sex (χ2 = 1.7702, p = 0.636), education level (F = 2.554, p _ANOVA = 0.065) and percentage of smokers (χ2 = 3.411, p = 0.332), drinkers (χ2 = 3.095, p = 0.377), hypertension patients (χ2 = 3.416, p = 0.3332) among the MDD (n = 14), PSND (n = 14), PSD (n =15) and CON (n =15) groups. The percentage of diabetic patients in the PSND and PSD groups was higher than in the CON and MDD groups (χ2 = 10.255, p = 0.017), but there was no significant statistical difference between the PSND and PSD groups. No significant difference was found in the NIHSS scores (t = -0.142, p = 0.888) between the PSND and PSD groups ( Figure 9B ). The HAMD-17 scores (F = 152.408, p _ANOVA < 0.001) of the MDD (MDD vs CON: p < 0.001, MDD vs PSND: p < 0.001) and PSD (PSD vs CON: p < 0.001, PSD vs PSND: p < 0.001) groups were significantly higher than those of the CON and PSND groups, but there was no significant statistical difference between the MDD and PSD groups ( Figure 9A , p = 0.685).

We examined the IL-7R expression in the plasm of MDD, PSD, PSND and CON groups ( Figure 9C , F = 30.053, p _ANOVA < 0.001). We found that the expression of IL7R was significantly lower in MDD (p _Turkey < 0.001), PSD (p _Turkey < 0.001) and PSND (p _Turkey < 0.001) groups than in the CON group. The expression of IL-7R in PSD group was lower than that in PSND group (p _Turkey = 0.041).